Volume 91, Issue 9 - Special Issue: Messunsicherheit, Datenqualität und datenbasierte Modellbildung/Measurement Uncertainty, Data Quality and Data-Driven Modelling

In situ calibration of sensors delivering SI traceable measurement results still provides an open question to the design and operation of sensor networks. Particularly when addressing low-cost sensors, currently, the use of sensor networks for air quality monitoring is limited by the low or unknown accuracy of measurements that they can achieve, while the data quality of individual sensor networks is mainly derived by algorithms. Standardization bodies like DIN and CEN therefore announced the need for investigations of validation methods on gas phase species and particulate matter on the one hand side, and for the development of fully digitized quality assurance/quality control and calibration techniques for sensor networks on the other (CEN/CENELEC, Opportunity for Standardisation to Contribute to the European Partnership on Metrology EPM under Horizon Europe ). This contribution concentrates on the metrological traceability of sensor networks for air quality monitoring to the international system of units (SI) based on FAIRified intra-network communications (M. Wilkinson, et al., “The FAIR guiding principles for scientific data management and stewardship,” Sci. Data , vol. 3, 2016, Art. no. 160018) and including delocalized Optical Gas Standards operated according to the digital TILSAM method (O. Werhahn, et al., The TILSAM Method Adapted into Optical Gas Standards – Complementing Gaseous Reference Materials , PTB Open Access Repository, 2021). Informed by related activities in EURAMET (Partnership project FunSNM, EMNs COO & POLMO, TC-IM 1551) (European Metrology Network Climate and Ocean Observation (COO), European Metrology Network Pollution Monitoring (POLMO), EURAMET Project TC-IM 1551, Project Database ) this contribution discusses the importance of measurement uncertainties in the context of sensor networks, comprising different sensor principles and promoting an efficient uptake of state-of-the-art methods. We discuss how the sensor network case can be addressed with sensors individually using the GUM principles (Joint Committee for Guides in Metrology, Guide to the Expression of Uncertainty in Measurement (GUM), JCGM 100: 2008 (E) ). For sensor network measurements becoming metrologically traceable to the SI, documented and unbroken chains of calibrations need to be implemented each contributing to the measurement uncertainty. This applies to each individual sensor of the network including the potential gold standard among them, but also to the network’s output viewed as a single entity. The contribution provides first approaches to be tested and validated that are underpinned by fundamental design strategies for sensor networks. It follows on practical applications in real world scenarios aside from model uncertainties discussed in artificial intelligence prospects.

This article describes the proof of concept of a Kibble extension for vacuum mass comparators and its further development. A Kibble extension is a technical extension for vacuum mass comparators to connect to the new definition of the International System of Units (SI) of 2019. A commercially available high-vacuum prototype mass comparator from Sartorius serves as the basis. It is used to compare kilogram prototypes, such as the international prototype kilogram ( K ) $(\mathfrak{K})$ , directly with other kilogram mass standards in order to realize a calibration chain. According to the new definition of the SI in 2019, the kilogram is defined by the Planck constant ( h ), therefore the Kibble extension is designed to enable high-vacuum mass comparators to be converted by technical modifications to allow mass to be realized according to the new definition. The aim is to realize kilogram prototypes and mass standards in the range from 1 mg to 1 kg according to the E1 standard of OIML R111-1 (2004). Initial measurement results with a prototype of the Kibble extension are presented and evaluated. Based on the experience gained, a lightweight magnet system developed with the use of simulation methods is presented. This is intended to contribute to the further development of the Kibble extension.

The analysis and reuse of measured process data are enablers for sustainable and resilient manufacturing in the future. Maintaining high measurement data quality is vital for maximising the usage and value of the data at hand. To ensure this data quality, the data management must be applied consequently throughout the complete Data Life-Cycle (DLC) and adhere to the FAIR guiding principles. In the two research consortia NFDI4Ing and the Cluster of Excellence “Internet of Production,” we investigate approaches to increase the measurement of data quality by integrating the FAIR guiding principles in all data management activities of the DLC. To facilitate the uptake of the FAIR guiding principles, we underline the significance of FAIR data for the reuse of high-quality data. Second, we are introducing a harmonised DLC to streamline data management activities. Third, we concisely review current trends and best practices in FAIR-aware data management and give suggestions for implementing the FAIR guiding principles.

Machine learning in industrial condition monitoring is currently a rapidly developing field of research, to improve the efficiency and reliability of industrial processes. Many of the used algorithms are supervised methods, which can learn and recognize hidden patterns in the data. However, training data is required to learn these patterns, which can only be generated to a limited extent in an industrial environment due to the high costs involved. Furthermore, it is impossible to represent all possible events in the training data. In contrast, unsupervised or semi-supervised methods can be used to detect new conditions or events. However, these usually do not allow diagnosis or quantification of a fault condition, which is why their usefulness for modern maintenance strategies is limited. Consequently, a robust condition monitoring system should combine the functionality of both approaches. This paper presents a methodology for the combination of supervised classification and semi-supervised novelty detection to build an expandable and adaptable condition monitoring by transferring recurring novelties as new conditions to the supervised classification. A superordinate algorithm is proposed to achieve a stepwise extension of the supervised model based on new conditions detected by novelty detection. With this approach, a condition monitoring system can at first be based on “normal” data of a new machine or process by adding failures or novel conditions step-by-step. Furthermore, the supervised methods can be used to help the corresponding staff identify unknown conditions by analyzing the features selected by the supervised classification. The general workflow is demonstrated for condition monitoring of the pneumatic drive system of a welding gun.

Systematic and random measurement errors are the cause of uncertain conformance and non-conformance statements. A wrong conformance statement occurs if the workpiece is accepted and is a reject part (type II error or false-negative) and a wrong non-conformance statement if the workpiece is rejected and is an in-spec part (type I error or false-positive). In order to avoid type I and type II errors, measurement uncertainty must be taken into account in the conformance and non-conformance testing. In practice, some procedures are used to consider measurement errors or the uncertainty that deviate from the state-of-the-art in research and technology. As these methods have become established over many years, they are still widely used despite better theoretical knowledge. The standard ISO 14253-1:2017 specifies a procedure based on probability and measurement uncertainty that is in accordance to the internationally accepted “Guide to the expression of uncertainty in measurement” and its supplements but is often not used due to the complexity of the evaluation of measurement uncertainty. In this contribution we give an overview and comparison of the different existing methods and provide an suggestion for supplementing the standard ISO 14253-1:2017, as Monte Carlo simulations enable a direct probability-based conformance and non-conformance testing even for complex measurement processes.

Mathematical models are vital to the field of metrology, playing a key role in the derivation of measurement results and the calculation of uncertainties from measurement data, informed by an understanding of the measurement process. These models generally represent the correlation between the quantity being measured and all other pertinent quantities. Such relationships are used to construct measurement systems that can interpret measurement data to generate conclusions and predictions about the measurement system itself. Classic models are typically analytical, built on fundamental physical principles. However, the rise of digital technology, expansive sensor networks, and high-performance computing hardware have led to a growing shift towards data-driven methodologies. This trend is especially prominent when dealing with large, intricate networked sensor systems in situations where there is limited expert understanding of the frequently changing real-world contexts. Here, we demonstrate the variety of opportunities that data-driven modeling presents, and how they have been already implemented in various real-world applications.